吸烟论文.docx

吸烟论文.docx

《吸烟论文.docx》由会员分享，可在线阅读，更多相关《吸烟论文.docx（11页珍藏版）》请在冰豆网上搜索。

吸烟论文

TheEmpiricalAnalysisof

theFactorsAffectingAdultSmoking

Abstract:

Byusingthe2006CHNScross-sectionaldata,thearticletriestofigureoutthemainfactorsinfluencingsmokingbehavior.ThemethodincludestwoIV,oneforincome（IVisnumberofstaffinjob’sorganization）andtheotherforhealth（IViswhethergettingmedicalinsurance）,aproxyvariable,dummyinteractionterm,LPMandlogitmodel.Theconclusionisageandhealthbothhavequadraticrelationshipwithsmoke,andthemostsignificantvariablesaremale,alcoholandeducation.

Student:

马超凡

IDnumber：

41110042

Ⅰ.Introduction

Nowadays,thenumberofsmokersinchinahasbeenover300million.Andtherearetotalabout540millionsmokersincludingpassivesmokers.Thecloserelationshipbetweensmokingandhealthcontributesmokingtobethefourthmostdangerousfactorthatinfluenceshealth.Everyyear,therewereabout1millionpeoplediedfromtobaccorelateddiseases.Thereisnodoubtthatsmokingreallybringsgrimchallengetohumansociety.

Smoking,asanaddictivebehavior,hascomplexcontributingfactors.Someofthemareevidentlikemale,education,alcohol,healthconditionandage.Someofthemjustexertsubtleinfluenceonsmokingbehaviorlikemarriedcondition,numberofsiblingsandsleep.Nowweareinaperiodofsocialtransition.Inthefastpaceoflife,thesefactorsareworkingtogethertoproducemoresmokers.

Tosolvethisseriousproblem,numerousrelevantresearcheshavebeenconducted.Mostresearchesareaboutthesmokingbehaviorsofteenagersandcollegestudents,fortheyarethemostpotentialsmokers.MaXiaobin,aprofessionaldoctor,findthatpeersmokingbehaviorsarecontagious.Onceastudentformsthisbadhabit,hewillattractandaffecthisfriendstoimitatehimandtrytosmoke.Andsubsequentlythisbadbehaviorcanbespreadwidely.Teenagersoftenhavenoawarenessofseriousharmofsmoking.Theyoncestartsmoking,thentheywillindulgeinitforthemoment’spleasure.Ontheotherhand,manystudentsjustregardsmokingasshowinguniquepersonality,whichencouragesthebadatmosphere.

XieJia,apostgraduatestudentofZhejiangUniversity,mainlyfocusesontheeffectsofmentalpressureonsmoking.Stressmeasurementresultsshowthatthemajoritypopulationisfacinggreaterstress,especiallythepeoplewithlowincome,loweducations.Therefore,effectivelysolvetheproblemofexcessivestresswillbeapowerfulfactortoreduceurbanresidents’smokingrate.Itissuggestedthatthegovernmentshouldfocusonthisgroupofpeople,notonlyconcernedabouttheirsmokingbehavior,butalsotocareaboutlivingandworkingpressure.

Afterreadingsomanypapersaboutsmoking,Ifindmostofthemareaboutteenagers.Eventhoughthereareafewpapersaboutadults,thesepapersarealmostaboutthesurveyresultsinmacroeconomics.Theyjustfailtodigoutthedeepreasonsbehindthesmokingbehaviors.Inmypaper,theresearchfocusesonadult.Firstfigureoutthecomprehensivecontributingfactorsofsmoking,andthenstudyhowtheyleadtosmoking.

Ⅱ.Datadescription:

Variabledescription:

1）Smoke（dependentvariable）:

thenumberofcigarettessmokedeveryday

2）Ifsmoke（dependentvariable）:

adummyvariable.“1”meansthepersonhassmoked.”0”meansthepersonneversmoked.

3）Education:

thenumberofyearsofeducationthepersonhasreceived.Iholdthebeliefthatpeoplewhoreceivedmoreeducationtendtosmokeless.

4）Male:

adummyvariablewith“1”representingmale.Ingeneralmenaremoreliketostartsmokeandsmokemorecigarettes.

5）City:

adummyvariable.“1”meansthepersonlivesincity.“0”meanshepersonlivesincountry.It’shardtosaywhichgroupsmokesmore.

6）Age:

theyearsoldtheperson

7）Sports:

adummyvariable.Definepeoplewhodon’tparticipateordislikewalking,taichi,allkindsofballgamesandfitnessactivitiesas“0”.Theotherisdefinedas“1”.

8）Income:

itisthequantityoftotalincomeinoneyear.Itiscalculatedbythefollowingequation:

（firstprofessionalmonthlywage+monthlysubsidy）*12+firstprofessionalyea-endbonus+（secondprofessionalmonthlywage+monthlysubsidy）*12+secondprofessionalyea-endbonus+farmingincome+Animalhusbandryincome+fishingincome+businessincome.

9）Starttime:

ifthepersonhassmoked,thisisthestartingageofsmoking.Maybestarttimeispositivelycorrelatedwithsmoke

10）Alcohol:

thisisadummyvariable.“1”meansthepersonwhodrinksalcohol.“0”meansthepersondoesn’tdrinkalcohol.Peoplewhodrinkalcoholaremorelikelysmoke.

11）Sleep:

thehoursthepersonspendsonsleepingperday.

12）Health:

thisistheresultofself-evaluation.“1”meansverygood,”2”meansgood,”3”meansordinary,and“4”meansbad.

13）Married:

adummyvariable.”1”meansthepersonhasmarriedandisstillhasamarriagerelationship.”0”meansthepersonwhoisunmarriedordivorcedorhisspousehasbeendead.

14）Siblings:

thenumberofsiblingsthepersonhas.

15）Midnsur:

adummyvariable,”1”meansthepersonhasamedicalinsurance.

16）Workers:

thenumberofstafftheperson’sjobfirmororganizationhave.

Possibleomittedvariables:

1）Thecigarette’sbrandandpriceeachsmokerbuys.Itisrelatedwithincomeandcityandmaybeeducation.Sothisomittedvariableisboundtoproducebiasandinconsistency.Weassumethatpeoplewhobuylowpricecigarettescansmokermoreeveryday,socigarettepriceispositivelyrelatedwithsmokingquantity.Thisomittedvariablehasapositivecoefficient.Thenweknowpeoplewhohavemoreincome,isincity,andhavehighereducationtendtobuyhigherpricecigarettes.Soincome,cityandeducationallhavepositivecorrelationswithcigaretteprice.Sothebiases（coefficientofprice*correlation）ofcoefficientsofthesethreevariables（income,cityandeducation）areallpositive.

2）Theworkingpressure.Becauseitisunobservable,Ifindaproxyvariable.

Whours:

theusualworkinghoursinoneweek.Inmyview,peoplewithmoreworkingpressuretendtosmokemore.Althoughhisproxyvariablecangeneratebiasonthreeparts:

intercept,workingpressureanderrorterm,itinsuresthattheallotherimportantvariablesareunbiased.Withoutthisproxyvariable,allexplanatoryvariableswillbebiasedbecauseofomittedvariable.

Sampleselection:

1）Ihavedroppedtheobservationswhoseanyvariableisnegative.

2）Dropalltheobservationswhosevaluesarebeyondtherangesetforeachvariable.

2）ThenIdeliberatelyhavedroppedtheobservationswhoseincome<100.Foronething,itisimpossibleforapersonhasincomeofaye1arlessthan100,sotheseobservationsareoutliers.Foranother,althoughthiscancausenonrandomsamples,ithasnoeffectonunbiasedandconsistentestimatorsinthepopulationmodel,fortheregressionfunctionE（smoke|income,age,maleandsoon）isthesameforanysubsetofthepopulation,whichisexogenoussampleselection.

Afterdoingthesetwosteps,thesamplesizeisreducedfrom9528to3179observations.

Ⅲ.Modelspecification:

Theoriginalmodelis

Smoke=β0+β1education+β2male+β3age+β4income+β5alcohol+β6married+β7siblings+β8sleep+β9city+β10sports+β11health+u

Next,itisneededtobemodified.

1）Firstly,nowthatfunctionalmisspecificationcanalsobeduetoomittedvariable,intheequation,whours（workinghoursperweek）asaproxyvariableofworkingpressureshouldbeadded.

2）Secondly,checkingwhetherquadraticorcubicitemsareneeded.

Iguessagehasaquadraticrelationshipwithsmoke,soIaddage^2（markedasage2）variable.Thisgraphdescribescigarettessmokedperdayonaveragefordifferentperiodsofage.Thedatausedareall3179observationsincluding2268nonsmokerswhosecigarettessmokedperdayare0.

Thegraphshowsthatsmokeisincreasingasageincreasesuntilabout50-55years-old.Thensmokedecreasesasageincreases.

UsingRESETtest,Iestimatetwomodelsforsmoke.Thefirstonehasallvariablesinlevelform.Thesecondoneaddsage^2.TheRESETstatisticforequation

（1）turnsouttobe26.7,thisisthevalueofanF（2,3130）,andtheassociatedp-valueis0.0000,thisistheevidenceoffunctionalmisspecificationin

（1）.TheRESETstatisticfor

（2）is24.6,withp-value=0.0000.Sowecanseetheequation

（2）isbetter,butthefunctionalmisspecificationisstillaproblem.

Thenusethe3significantvariables’square（health,education,siblings）respectivelyaddedtotheequationtoseewhetherthefunctionisimproved.Thenhealthsquaregivesasignificantt=-2.96.Soremainingthehealthsquareintotheequationisnecessary.

3）Thirdly,checkwhetherlogitemsisneededwith.Incomeshouldbeusedasln（income）,becauseincomerangesfrom100to960,720yuan.Soweshouldtrytoshortentherange.Afterdoingthis,incomerangesfrom4.60517to13.77544.AndtheMizon&Richasrdtestalsoshowslnincomeisbetter.Itrytouseln（whours）toreplacewhours,but756observationshave0workinghours,solnwhoursisnotavailable.

4）Fourthly,usechow-testtocheckwhetherinteractiontermsareneeded.Atfirst,Iusemarried*other12variablesrespectivelytoaddtothefunction.ThenF（12,3116）=1.46withp-value0.13,sowecan’trejectthenullhypothesisat5%significancelevelandevenat10%significancelevel.Sotheinteractiontermsarejointlyinsignificant.

ThenIusemale*other12variablesrespectivelytoaddtothefunction.TheF=7.2withp-value0.0000.Sothereisnodoubtthatsomemaleinteractiontermsmustbeadded.ThenIchoosefourverysignificantinteractiontermsmale*education,male*age,male*age2andmale*alcohol,witht-statisticsareover5toaddtotheequation.Unfortunately,originallysignificantvariableslikeeducation,male,age,